| SOLVED: Hang on vzctl start/stop in futex() in the ve. [message #29159] |
Mon, 07 April 2008 15:19 |
vidarh
Messages: 1
Registered: April 2008
|
Junior Member |
|
|
The problem was that /etc/nsswitch.conf had "ldap" listed before "compat" or "files" for a number of entries and this was the first time the VE's in question had been restarted. Apparently bash locks up hard in that case if it can't reach the network, and since the scripts to add the IP address runs under bash inside the VE, well, it didn't work too well.
For us it's ok to just switch and put ldap last. If that doesn't work for someone else, the solution would seem to involve switching nsswitch.conf files around on boot/shutdown
---
At some point today, one of my containers got unresponsive, and when strace'ing the processes involved I noticed it was hanging in the "futex" syscall.
I then tried vzctl stop, but the stop took ages. When looking, "shutdown" was hanging in the futex syscall. I waited for it to time out, and restarted again. Same thing.
I then tried to use vzdump, followed by vzctl destroy and vzdump --restore. Same thing again when I tried vzctl start.
UPDATE: I also tried to stop/start OpenVz with /etc/init.d/vz restart
I'll reboot the system as soon as I can, but I'd like to know if there's a way to fix this without a reboot if possible - it's a bit of a concern to me if we risk having to reboot production machines because of problems with a single VE.
Any ideas?
(The kernel used is 2.6.18-fza-028stab053.5-686)
[Updated on: Tue, 08 April 2008 10:15] Report message to a moderator |
|
|
|
OpenVZ Forum
Members
Search
Help
Register
Login
Home
